Generation Jobs

Generation requests are asynchronous. The trigger request returns quickly, then the backend continues processing the job.

The common pattern is:

Start work with POST /images/generation, POST /images/upscaling, or POST /videos/generation.
Save the returned jobId.
Poll GET /jobs/{id} until the status leaves the running state.
Read generated images, generated videos, or error from the job payload.

Training note

The OpenAPI schema documents GET /models/{modelID}/train as returning a success string, not a jobId. For training, track readiness by polling the model with GET /models/{modelID} until status becomes ready. Training usually takes 20 to 25 minutes, so poll about every 5 minutes instead of every second.

Job response

GET /jobs/{id} returns a result object with job metadata and any completed outputs.

Field	Description
`jobId`	Job identifier returned by the trigger request.
`status`	Current backend state.
`images`	Generated image records when the job produced images.
`videos`	Generated video records when the job produced videos.
`error`	Failure reason when the job did not complete.
`modelId`	Related model ID when the job is model-scoped.
`createdAt`	Job creation timestamp.

Poll a job

Use a short delay between polling attempts. The current job status values are created, queued, running, success, and failed.

const RUNNING_STATUSES = new Set(['created', 'queued', 'running'])
const SUCCESS_STATUS = 'success'

async function waitForJob(jobId, intervalMs = 5000) {
  while (true) {
    const response = await fetch(`${BASE_URL}/jobs/${jobId}`, {
      headers,
    })

    if (!response.ok) {
      throw new Error(await response.text())
    }

    const body = await response.json()
    const job = body.result
    const status = String(job.status ?? '').toLowerCase()

    if (!RUNNING_STATUSES.has(status)) {
      return job
    }

    await new Promise((resolve) => setTimeout(resolve, intervalMs))
  }
}

const job = await waitForJob('<JOB_ID>')
const status = String(job.status ?? '').toLowerCase()

if (status !== SUCCESS_STATUS) {
  throw new Error(job.error ?? `Job failed: ${job.status}`)
}

import time

RUNNING_STATUSES = {"created", "queued", "running"}
SUCCESS_STATUS = "success"


def wait_for_job(job_id: str, interval_seconds: int = 5):
    while True:
        response = requests.get(f"{BASE_URL}/jobs/{job_id}", headers=headers)
        response.raise_for_status()

        job = response.json()["result"]
        status = job.get("status", "").lower()

        if status not in RUNNING_STATUSES:
            return job

        time.sleep(interval_seconds)


job = wait_for_job("<JOB_ID>")

if job.get("status", "").lower() != SUCCESS_STATUS:
    raise RuntimeError(job.get("error") or f"Job failed: {job.get('status')}")

Read outputs

Image generation and upscaling jobs return image records in images.

for (const image of job.images ?? []) {
  console.log(image.id, image.url)
}

for image in job.get("images", []):
    print(image["id"], image["url"])

Video generation jobs return video records in videos.

for (const video of job.videos ?? []) {
  console.log(video.id, video.url)
}

for video in job.get("videos", []):
    print(video["id"], video["url"])

Generation Jobs

Job response

Poll a job

Read outputs

API reference

Get job status

Generate images

Upscale images

Generate videos

On this page